Moore's Law Is Almost Broken | Dr Sung-hyun Park, Rebellions

Not yet. Almost. Dr Sung-hyun Park on why the industry now measures TOPS instead of transistors. Book Ian for your event → iankhan.com | booking@iankhan.com SUMMARY Dr Sung-hyun Park founded Rebellions to build specialised AI chips. He explains why training and inference are genuinely different workloads, why each chip should be specialised to a service and an algorithm, and why the metric the industry quotes has already changed. KEY FINDINGS • AI requiring parallel computation where CPUs handled sequential work • Training and inference as meaningfully different workloads • Chips specialised for a specific service and algorithm rather than general purpose • Moore's law almost, but not quite, broken • The industry shifting to TOPS rather than transistor count • What that slow decline means for chip design • Where specialisation goes over the next two to five years NOTABLE QUOTES "Moore's law is almost broken. Not yet, but almost." — Dr Sung-hyun Park GUEST Dr. Sung-hyun Park, Founder & CEO, Rebellions Inc. FULL TRANSCRIPT iankhan.com/futurist/sung-hyun-park-rebellions-ai-chips Recorded at the Future Investment Initiative, Riyadh, for The Futurist. ──────────────────── ABOUT IAN KHAN — AI KEYNOTE SPEAKER Most AI keynotes end with a better understanding. Mine end with a measured readiness score and a ninety-day plan. Your leadership team is assessed before the event, the room sees its own collective result live, and every attendee leaves with a ninety-day plan they helped build.